Date: Sun, 4 Jan 2015 20:06:37 +0100 From: Mateusz Guzik <mjguzik@gmail.com> To: "Bjoern A. Zeeb" <bz@FreeBSD.org> Cc: svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r276424 - head/sys/kern Message-ID: <20150104190637.GA5807@dft-labs.eu> In-Reply-To: <F9979C81-4F4C-4623-B0EE-FA11F8011D7B@FreeBSD.org> References: <201412302140.sBULek5f076019@svn.freebsd.org> <F9979C81-4F4C-4623-B0EE-FA11F8011D7B@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Dec 30, 2014 at 10:45:43PM +0000, Bjoern A. Zeeb wrote: > > > On 30 Dec 2014, at 21:40 , Mateusz Guzik <mjg@FreeBSD.org> wrote: > > > > Author: mjg > > Date: Tue Dec 30 21:40:45 2014 > > New Revision: 276424 > > URL: https://svnweb.freebsd.org/changeset/base/276424 > > > > Log: > > Convert vfs hash lock from a mutex to an rwlock. > > This is very obvious from the change. > > What is not and should be described as well is some of this: > > - Why? > - What are the benefits? > - What are the locking implications? > - Does it make a performance difference? If yes, was it measured? > - … (you got the idea) > Sorry for late reply. This is just one of many minor things which showed up in lock profiling during make -j 40 buildkernel. I don't expect the gain to be quite noticeable (at least not now since there way more serious bottlenecks), thus there are no claims stuff is way faster in the commit message. This change is entirely transparent to callers and callees. -- Mateusz Guzik <mjguzik gmail.com>
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20150104190637.GA5807>